swept blend acting crazy

[2ms1]

Shown below is a swept blend with 11 cross sections that are each positioned by relations. The relations dictate the length of the trajectory curve, the angle the cross sections form relative to perpendicular to the traj curve, and the rotation ("twist") of the sections about the traj curve.*


The big problem that I have spentforever trying to resolve with no progress, is that half the time I change any of these parameters the sweep goes crazy. See below for example. Specically, the cross section planes seem to always want to flip 180°, and even more strangely 90deg sometimes.I'm also pasting the entire set of relations for the sweep below. You can see that Iinitially,the first time it happened,attempted to fudge corrections for the flips by simply doing things like adding 180, -90, etc. to the dimensions that determine the rotation of these sections, but these corrections have turned out to be useless because they'll go crazy again the next time I change a parameter value.


I know this is long but Id really appreciate it if someone could look at this and see if you have any ideas.


I understand the relations have a lot of variables that you wouldn't know by name, but any of those wouldn't be connected to this crazy being rotated ("twist") about the trajectory by 90,180, etc degrees. The problems lie with the section commented by: " /* twist "


View attachment 4055


crazy:


View attachment 4056


/*----------------------------------------
/*----------------------------------------
/* Distance between joint centers (length of trajectory curve)
Zdist=231



/* Width of tipside joint
maxwtipside=60



/* Height of tipside joint
maxhtipside=51



/* Width of palmside joint
maxwpalmside=65.5



/* Height of palmside joint
maxhpalmside=52



/* Width center of segment
minw=35



/* Height center of segment
minh=34



/*Twist
twist=37.38


delx=50
dely=20


/*----------------------------------------
/*----------------------------------------
jointoffset=1*(17-5)


/*twistadjustx=(maxhpalmside/2)*sin(twist)
/*twistadjusty=(maxhpalmside/2)*(1-cos(twist))


q=twist
halfmaxh=maxhpalmside/2
sqrtterm=abs(sqrt((halfmaxh*tan(q))^2)-(halfmaxh^2))
s=sqrtterm-halfmaxh
L=sqrt((halfmaxh^2)+(halfmaxh+s)^2)
twistadjustx=L-(s*L/(halfmaxh+s))





/*twistadjustx=(
twistadjusty=halfmaxh*(sqrtterm-halfmaxh)/sqrtterm


thetax=atan((delx-twistadjustx)/(Zdist))
thetay=abs(atan((dely+twistadjusty)/(Zdist)))


hypotenusex=((Zdist^2)+((delx+twistadjustx)^2))^(1/2)
hypotenusey=((Zdist^2)+((dely-twistadjusty)^2))^(1/2)


abslength=Zdist+(hypotenusex-Zdist)+(hypotenusey-Zdist)



psi=1
psiadjusted=psi


Curvaturey=seglength/(2*sin((psiadjusted)/2))
d601=Curvaturey


d670=seglength
d734=d670


/* twist
$d747=-(twist+90)
$d767=-(90+twist)
$d784=-(.87*twist+90)
$d793=-(.58*twist+90)
$d802=90-.54*twist
$d812=-90-.5*twist
$d820=90-.46*twist
$d827=90-.42*twist
$d837=-90-.13*twist






/* section datum plane rotations about X


angle1=(psiadjusted/2-thetay/2)
angle9=(psiadjusted/2-thetay/2)
angle2=angle1-(.08*angle1)
angle3=-(angle1-(.15*angle1))
angle4=-(angle1-(.95*angle1))
angle45=0
angle5=0
angle55=0
angle6=angle9-.90*angle9
angle7=-(angle9-.15*angle9)
angle8=angle9


$d646=angle1
$d659=angle2
$d669=angle3
$d649=angle4
$d650=angle45
$d651=angle5
$d848=angle55
$d795=angle6
$d785=angle7
$d667=angle8
$d658=angle9





/* section datum plane z-positions


zsec1=0
zsec9=seglength


zsec2=.04*zsec9
zsec3=.15*zsec9
zsec4=.42*zsec9
zsec45=.46*zsec9
zsec5=.5*zsec9
zsec55=.54*zsec9
zsec6=.58*zsec9
zsec7=.85*zsec9
zsec8=.96*zsec9



d332=zsec2
d336=zsec3
d339=zsec4
d342=zsec45
d345=zsec5
d348=zsec55
d351=zsec6
d354=zsec7
d357=zsec8
d360=zsec9



/* section sketch y-offset from trajectory curve


hdt=maxhtipside-minh
hdp=maxhpalmside-minh


ysec1=0
ysec2=0
ysec3=.06*hdt
ysec4=.12*hdt
ysec45=ysec4
ysec5=ysec45
ysec55=ysec5
ysec6=.12*hdp
ysec7=.06*hdp
ysec8=0
ysec9=0



d665=ysec1
d375=ysec2
d874=ysec3


d879=ysec4
d825=ysec45
d889=ysec5
d854=ysec55
d916=ysec6
d790=ysec7
d894=ysec8
d759=ysec9



/* sketch width


wsec1=maxwtipside
wsec2=wsec1
wsec3=wsec1-.2*(maxwtipside-minw)
wsec4=minw
wsec45=minw
wsec5=minw
wsec55=minw
wsec6=minw
wsec7=maxwpalmside-.2*(maxwpalmside-minw)
wsec8=maxwpalmside
wsec9=wsec8


d365=wsec1
d373=wsec2
d871=wsec3
d876=wsec4
d822=wsec45
d886=wsec5
d851=wsec55
d915=wsec6
d786=wsec7
d890=wsec8
d755=wsec9



/* sketch height


hsec1=maxhtipside
hsec2=hsec1
hsec3=hsec1-.2*(maxhtipside-minh)
hsec4=minh
hsec45=minh
hsec5=minh
hsec55=minh
hsec6=minh
hsec7=maxhpalmside-.2*(maxhpalmside-minh)
hsec8=maxhpalmside
hsec9=hsec8


d366=hsec1
d374=hsec2
d872=hsec3
d877=hsec4
d823=hsec45
d887=hsec5
d852=hsec55
d916=hsec6
d787=hsec7
d891=hsec8
d756=hsec9



/* tendon indent outward shift
d368=0
d372=0
d873=0
d878=2.4
d824=2.4
d888=2.4
d853=2.4
d917=2.4
d789=0
d893=0
d758=0





*There are also a bunch of relations for the width and height of each sketch on each cross section plane, the spacing of them, etc but these are not causing problems -- the ones related to the sweep itself (the ones listed in previous paragraph) are the ones causing problems.
Edited by: 2ms1

 

[puppet]

id say its something simple.
goto those sections and make sure ur starting vertex is in the same location as all other previous sections

 

[gkbeer]

For a feature like this it pays to be absolutely meticulous that every section is setup/created in the exact same way as the rest.

 

[wsylvester]

itdoes look like it's sync


if that's variations of the same section you'd could get away from the sync problem by using evalgraph and drive the section changes and twistwith the graphs using your relations.

 

[2ms1]

what do you mean when you say it does look like it's sync exactly?


It is variations of the same section. I just looked up what evalgraph is in Pro/E's help, and I get the basic idea of it, butpleaseelaborate on your idea of using it to drive the section changes and twist.
Edited by: 2ms1

 

[wsylvester]

It the same thing that puppet and glenn are talking about.


but to evalgraph..


you will need to have all your drive curve relations anything to do with the spine first.. so the curve is complete..okay


then create a graph..take the default name.. then when it come up drop a local csys some cl's for your axis and values.


start a line at say ..x0 y 10 at x10y20 and end at x20y10


It might help to think of a evalgraph asa function of time, with x axisbeing time no twoy values can be at the same place in time.


since you want to put twist on your section you need to place your sketch so it is angularly constrained first off.... you can build on this as you go along.


so you have your section all place properly... now toggle dims so you can see what the sketch dim is for the angle..if you drew the angle line first it will more than likely be sd4....


so while still in sketch > tools>relations add;


sd4=evalgraph("graph1",trajpar*20)


note the *20 is the same as the x units in the graph, if you use *10, graph being 20 you will only eval half the graph.


...so it okay, complete feature and it should start at 10 degrees twist, midpoint will be 20 degrees twist, end back to 10.


then after getting all the twist worked out to the proper place then you can go on to the changes in width and height of the section by creating the graphs needed and then dragging them above the vss applying the evalgraph relation to the proper sketch dim.by edit defination... goto the sketch and then add inevalgraph relation..


you should be able to bag some of your existing relations into the graph and apply those values there.


I think once you got the twist worked out it will be more clearer, play around with it see if it will fit your need... I think it will

 

[2ms1]

Ok, I will have to spend some time thinking about this evalgraph approach.


Before I commence with the giant overhaul though, I'd like to have some idea what is flawed about my current one. The current one seems so natural, valid,and fundamentally straight-forward. I'd really like to have some idea how it went wrong. When I look at the definition of the blend itself, all the sections have their vertexes at same point in each sketch, just that the one"crazy"section is rotated 180deg. Is this what you are talking about when you say sync?


Any other ideas what is flawed about this approach relative to the evalgraph approach? I'm concerned if I just remove the current dimensions and replace them with new approach dimensions I might still have the same problem.
Edited by: 2ms1

 

[wsylvester]

My best guess it has something to do with Intent Manager., but then that a guess.


The evalgraph method may seem like a giant overhaul but it allows you to chip away at it... get your twist.. then get your width.. height...etc...


if you use your relation valuesin the graph you'll be able to see what if any problems are because it will be quite noticeable...


You may need to create a analyse feature for your spine len and calc some ratios for your graph features x length.. because it is a function of it's travel down the traj.


You'll have more sections , but they're easy to add once you get started


View attachment 4066